【作者】 方兆宝；

【导师】 王金庄； 吴立新； 林珲；

【作者基本信息】 中国矿业大学（北京） ， 摄影测量与遥感， 2004， 博士

【摘要】 随着空间信息获取技术的进步,现代数据采集技术得到了飞速的发展,不同类型的与气象有关的空间数据库以及数据库中的数据量正在快速增长,但大量的、有用的信息也被淹没在这些数据之中,没有得到充分的开发和利用。如何从这些海量数据中提取与灾害天气有关的模式、模型,进而成为预测灾害天气发生的信息和知识,是当前遥感学、地理信息系统、气象学家研究的前沿、重点和热点。但由于气象数据的多数据源、多类型、多时相、多分辨率等固有的特点,指望一个数据挖掘系统挖掘所有类型的气象数据是不现实的。因此,为挖掘特定类型的数据,应当构造特定的数据挖掘系统。本文针对气象数据类型的多样性、复杂性,以及空间数据挖掘在气象中的应用存在的技术问题,以空间数据挖掘技术在青藏高原中尺度对流系统(MCSs)时空演变规律研究为背景,研究解决了如下问题:1.如何从气象卫星红外遥感影像中定量化及客观地提取MCSs空间特征,构造特殊的MCSs时空数据库。2.如何构造MCSs空间数据挖掘数据库3.寻求有效的数据库约简方法4. MCSs时空演变规则挖掘及其知识发现的可视化表达本文将空间数据挖掘技术应用于青藏高原MCSs时空演变规律研究,对MCSs的移动和传播这一迄今的难题提供了研究思路和方法,其挖掘结果对预报高原MCSs东移影响长江中下游地区的强降水也很有帮助。更多还原

【Abstract】 With the advancement of spatial information acquisition technology, modern data acquisition technology develops rapidly. Different types of spatial database and the data quantity that is relative with weather is increasing fast, but large amount of useful information is submerged in these data and can not got the best exploration and use. How to pick up the mode and model that is relative to disaster weather from these mass data and take them as the information and knowledge to forecast the happening of disaster weather become the researching emphasis, foreland and hotspot of remote sensing, geographical information system and meteorology.But due to the inherent characteristics of the multi data source, multi type, multi temporal and multi resolution, one data mining system to excavate all types of weather data is unrealistic. So, in order to excavate specific types of data, specific data mining system should be constructed. Aiming at the diversity of weather data types and the technique problems existing in the application of spatial data mining, this thesis has a research and resolves the following problems, with the background of the research of Mesoscale Convective Systems (MCSs) Spatio-temporal evolution rules in Tibetan Plateau using spatial data mining technology.1. How to extract the spatial features of MCSs in fit quantity and impersonal from the weather satellite infrared remote sensing image and construct specific MCSs spatial database?2. How to construct MCSs spatial data mining database.3. Seek efficient database reduction methods.4. MCSs spatial evolution rules mining and its visual expression of knowledge discovery.The data mining on MCSs in this thesis provides the research thought and methods for the study on moving and propagation of MCSs. The mining result is also helpful to forecast heavy rainfall in Yangtze River Basin with forecasting MCSs moving out to east from Tibetan Plateau.更多还原